Ensure worktree execution starts only after activation (#9374)

## Thinking Path

> - Paperclip is the open-source control plane people use to manage AI
agents and their work.
> - Its scheduler, routines, and heartbeat services decide when agents
automatically begin work.
> - Experimental per-worktree execution is useful for isolated
development, but enabling it previously allowed automatic services to
consider an existing backlog.
> - A worktree activation must therefore create a durable eligibility
boundary rather than merely toggle execution on.
> - This pull request records an activation cutoff and applies it
consistently to automatic routine and heartbeat dispatch.
> - The result is that an enabled worktree executes only work created
after its own activation, while non-worktree behavior remains unchanged.

## Linked Issues or Issue Description

**Problem type:** Bug / safety regression

**Summary:** Enabling experimental run execution in an existing worktree
could start automatic scheduler, routine, watchdog, and heartbeat
activity for work created before that worktree was explicitly armed.

**Expected behavior:** A worktree that has execution enabled only
considers automatically dispatched work created on or after its
activation timestamp. Ambiguous activation state fails closed.
Non-worktree instances keep their existing behavior.

**Related public work:** Refs #8275 (runtime worktree policy gating);
this PR adds an activation-time boundary for automatic execution rather
than changing the general runtime policy.

## What Changed

- Persist a worktree execution activation timestamp and originating
instance ID; stamp them only when the experimental toggle changes from
disabled to enabled.
- Resolve activation state fail-closed when the cutoff is missing,
invalid, disabled, or belongs to another instance.
- Gate automatic routine scheduling, webhooks, watchdog activity, and
heartbeat selection at the activation cutoff; manual runs remain
available.
- Share the canonical worktree truthy-environment helper across routine
dispatch and agent inbox filtering.
- Add cutoff and truthy-runtime regression coverage, plus
experimental-settings UI states that explain armed and suppressed
execution.

## Verification

- `pnpm exec vitest run server/src/__tests__/routines-service.test.ts
server/src/__tests__/instance-settings-service.test.ts` — passes: 2
files, 60 tests.
- `pnpm --filter @paperclipai/server typecheck` — passes.
- Existing CI completed successfully before the follow-up review fixes;
this branch was rebased onto the latest `origin/master` before
retesting.

## Risks

- **Behavioral:** Automatic worktree execution is intentionally more
restrictive; pre-existing work is suppressed until newly created after
activation.
- **Operational:** A malformed or cross-instance activation record fails
closed, requiring an operator to disable and re-enable the experimental
toggle on the intended worktree.
- **Compatibility:** The worktree environment now accepts all canonical
truthy values (`1`, `true`, `yes`, and `on`) consistently; non-worktree
instances are unaffected.
- **Branch metadata:** This existing execution-workspace branch predates
the current naming rule and cannot be renamed under this task's
workspace contract; the code and PR title do not include internal ticket
references.

> `ROADMAP.md` was checked; this targeted execution-safety fix does not
duplicate planned core work.

## Model Used

- Anthropic Claude Code — assisted with the original implementation;
exact model identifier and context window were not recorded in the
repository metadata.
- OpenAI Codex CLI — assisted with PR preparation and review fixes;
exact model identifier and context window are not exposed in this
execution environment. Used with terminal tooling, code editing, and
targeted test execution.

## Checklist

- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [ ] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [ ] All Paperclip CI gates are green
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

---------

Co-authored-by: Paperclip <noreply@paperclip.ing>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

packages/shared/src/types/instance.ts

@@ -70,6 +70,16 @@ export interface InstanceExperimentalSettings {
    * runs actually execute inside the preview. Ignored outside a worktree.
    */
   enableWorktreeRunExecution: boolean;
+  /**
+   * Server-managed cutoff recorded when worktree run execution is enabled in
+   * this instance. Client PATCH payloads must not control this value.
+   */
+  worktreeRunExecutionActivatedAt: string | null;
+  /**
+   * Server-managed instance id captured with the cutoff so copied settings rows
+   * from another instance fail closed.
+   */
+  worktreeRunExecutionActivationInstanceId: string | null;
   issueGraphLivenessAutoRecoveryLookbackHours: number;
 }
 

packages/shared/src/validators/instance.test.ts

@@ -28,6 +28,20 @@ describe("instance experimental settings validators", () => {
     const settings = instanceExperimentalSettingsSchema.parse({});
 
     expect(settings.enableWorktreeRunExecution).toBe(false);
+    expect(settings.worktreeRunExecutionActivatedAt).toBeNull();
+    expect(settings.worktreeRunExecutionActivationInstanceId).toBeNull();
+  });
+
+  it("strips server-managed worktree run execution fields from patches", () => {
+    expect(
+      patchInstanceExperimentalSettingsSchema.parse({
+        enableWorktreeRunExecution: true,
+        worktreeRunExecutionActivatedAt: "2026-07-10T12:00:00.000Z",
+        worktreeRunExecutionActivationInstanceId: "copied-instance",
+      }),
+    ).toEqual({
+      enableWorktreeRunExecution: true,
+    });
   });
 
   it("defaults built-in agents off", () => {

packages/shared/src/validators/instance.ts

@@ -58,6 +58,8 @@ export const instanceExperimentalSettingsSchema = z.object({
   enableWorkspaceBranchReconcileForward: z.boolean().default(true),
   enableWorkspaceDirtyQuarantineRepair: z.boolean().default(true),
   enableWorktreeRunExecution: z.boolean().default(false),
+  worktreeRunExecutionActivatedAt: z.string().datetime().nullable().default(null),
+  worktreeRunExecutionActivationInstanceId: z.string().min(1).nullable().default(null),
   issueGraphLivenessAutoRecoveryLookbackHours: z
     .number()
     .int()
@@ -66,7 +68,13 @@ export const instanceExperimentalSettingsSchema = z.object({
     .default(DEFAULT_ISSUE_GRAPH_LIVENESS_AUTO_RECOVERY_LOOKBACK_HOURS),
 }).strict();
 
-export const patchInstanceExperimentalSettingsSchema = instanceExperimentalSettingsSchema.partial();
+export const patchInstanceExperimentalSettingsSchema = instanceExperimentalSettingsSchema
+  .omit({
+    worktreeRunExecutionActivatedAt: true,
+    worktreeRunExecutionActivationInstanceId: true,
+  })
+  .partial()
+  .strip();
 
 export const patchInstanceSettingsSchema = z.object({
   defaultEnvironmentId: z.string().uuid().nullable().optional(),